Using correlation structure of speech dynamics to detect neurological changes

1. A computer implemented method of assessing a condition of a subject, the method comprising: receiving, at a computing device, a digitized microphone signal representing an acoustic signal, including the subject's speech, received at a microphone of the device; processing, using the computing device, the digitized microphone signal to produce successive values of at least one speech-related variable; determining, using the computing device, a plurality of delays of the at least one speech-related variable; calculating, using the computing device, a channel-delay correlation or covariance matrix from the plurality of delays of the at least one speech-related variable; determining, using the computing device, a correlation structure of the at least one speech-related variable, including determining a matrix eigenspectrum from the channel-delay correlation or covariance matrix; generating, using the computing device, an assessment of the condition of the subject, based at least in part on the matrix eigenspectrum of the correlation or covariance matrix of the correlation structure of the at least one speech-related variable; and displaying, on a display, the assessment of the condition of the subject for use by a clinician to predicted, diagnosed, or monitor the condition of the subject.

2. The method of claim 1 , wherein the at least one speech-related variable includes a formant frequency.

3. The method of claim 2 , wherein the at least one speech-related variable includes two or more formant frequencies.

4. The method of claim 1 , wherein the at least one speech-related variable includes a facial action unit, the facial action unit corresponding to muscle movements of the face.

5. The method of claim 1 , wherein the at least one speech-related variable includes a Mel Frequency Cepstral Coefficient.

6. The method of claim 1 , wherein the at least one speech-related variable includes a Delta Mel Frequency Cepstral Coefficient.

7. The method of claim 6 , wherein the speech-related variables are two or more Delta Mel Frequency Cepstral Coefficients.

8. The method of claim 1 , wherein the channel-delay correlation or covariance matrix includes channel-delay correlation values.

9. The method of claim 1 , wherein the channel-delay correlation or covariance matrix includes channel-delay covariance values.

10. The method of claim 1 , wherein the condition is selected from traumatic brain injury, post-traumatic stress disorder, Parkinson's disease, Aphasia, Dysphonia, Autism, Alzheimer's disease, Amyotrophic Lateral Sclerosis, often referred to as Lou Gehrig's Disease, stroke, sleep disorders, anxiety disorders, multiple sclerosis, cerebral palsy, and major depressive disorder.

11. The method of claim 1 , wherein the condition is major depressive disorder.

12. The method of claim 1 , wherein: the at least one speech-related variable are includes the first three formant frequencies; and the condition is major depressive disorder.

13. The method of claim 1 , wherein: the at least one speech-related variable includes the first sixteen Delta Mel Frequency Cepstral Coefficients; and the condition is major depressive disorder.

14. The method of claim 1 , wherein the condition is major depressive disorder, and wherein generating the assessment of the condition includes generating an estimate of a Beck score of the subject, an estimate of a Hamilton-Depression score of the subject, or an estimate of a Quick Inventory of Depressive Symptomatology score of a subject.

15. The method of claim 14 , further including displaying the estimate of the Beck score, the Hamilton-Depression score or a Quick Inventory of Depressive Symptomatology score of the subject.

16. The method of claim 1 , wherein processing the digitized microphone signal comprises determining a vocal tract representation, and wherein generating the assessment of the condition of the subject is based at least in part on time correlation structure of the vocal tract representation.